Chatgpt-4o was a marvel for exactly 1wk after relase… it has got so bad really bad these days that it has become a challenge not to smash the keyboard. I’ll break it down below:
Time Waste: It is so bad at coding even when i clearly indicate what documentation to account for that i am spending more time correcting it.
Comprehension: it has no understanding of what i ask of it… or it simply chooses to answer another question that i haven’t asked.
Verbosity: I have added all over the place that i want short answers, in the gpt context, created my own gpt i have did everything i can… It will not stop vomiting pages and pages of garbage i don’t care about. I tell it at the beginning of the conversation and after 3 messages completely disregards it.
Repeating itself: The most important thing about LLM’s is the context… why would it redo 50 lines of code to change a param at the beginning… why can;t it just take a block of code that is in the discussion… and highlight add or modify instead of redoing the whole thing… then it burns trough the tokens and it becomes even more useless.
Won’t read docs: I tell it from the first chat that it does not have the necessary knowledge, and it has to consult X documentation… it chooses not to and then vomits whatever it thinks the solution is.
Critical Thinking: dead… nothing else to add.
I am willing to pay $$$ can i not get the good version i’ll pay for my compute and then some for profit… this is ridiculous… I think i will switch as soon as Claude becomes better… now it’s the same but did not receive 10 Billion USD … they did it on pocket money.